Deciding and acting on quality of microarray experiments in genomics

1
Deciding and acting on quality of microarray
experiments in genomics

• Chris EveloBiGCaT Bioinformatics Maastricht

2

Gene Expression
From Alberts et al. Molecular Biology of the
Cell, 3rd edn.
3

First Example
Is red wine healthy?
Does it protect rats from eating the unhealthy
stuff we usually eat?
4
Experimental design

• Control group10 male F344 ratsDiet high fat
  (23), high sucrose, low fibre
• Experimental group 10 male F344 ratsSame diet
  plus 50 mg/kg red wine polyphenols

10 treated 50 mg/kgday, 2 wks Cy 5
Pool of 10 controls Cy 3
5
DNA Microarray
6
Microarray Principle
7
The genomics workflow
8
Before our analysis

• Conclusions disagree with previous results
• 690 genes regulated genes
• Involved incell adhesion and cell-cell
  communication
• Instead ofe.g. antioxidant activity

9
Quality control-using Spotfire DecisionSite- (I)
Microarray laser scan. 16 Print blocks
Created with Spotfire DecisionSite Colors
represent feature numbers of spots on microarray
10
(No Transcript)
11
Quality control-using Spotfire DecisionSite- (II)

• Localization of the flagged features (empty spots
  and bad spots (e.g. Signal lt BG))
• Flagged features are removed for further analysis

12
Hierarchical Clustering
13
K-means Clustering
14
Dissimilar Genes
690 genes
15
Dissimilar Genes
16
(No Transcript)
17
Disagreement with biological data
?
18
Questions

• Differences due to the dietary treatment?
• Check on the rats growth during the experimental
  time and on their
• weight at sacrifice
• Differences due to the natural inter-individual
  variability?
• Fischer 344 are inbred rats, genetically very
  similar. A variability among rats is (of course)
  possible but unlikely in this case, due to the
  type of treatment and to the large amount of
  differences observed (more than 600 genes
  differentially expressed)
• Technical problem?

19
Localization of the differentially expressed
genes-using Spotfire DecisionSite-
20
Log ratio
21
(No Transcript)
22
(No Transcript)
23
Visualize expression results
SwissProt
24
Most important results of genMAPP
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30

Conclusions

• Using Spotfire Decisionsite we can
• see problems on microarrays
• see unexpected things using variable sliders
• group co-expressed genes (clustering, pca)
• see the location of specific genes or groups of
  genes
• immediately see the effects of alternative
  treatments
• combine with biological interpretation in
  GenMAPP

31
Example 2 Antibody MicroarrayBD Biosciences
(Clontech)

• Chip-based technology
• Monoclonal antibodies printed at high density on
  a glass slide
• Profiling hundreds of proteins
• Analyses virtually any biological sample (cells,
  whole tissue and body fluids)

32
Content of antibody array
33
(No Transcript)
34
Two slides with flipped samples
35
Internally normalized results

• Sampling method controls for differences in
  labeling efficiency
• Internally Normalized Ratio can be calculated
• (represents the relative abundance of an
  antigen in sample A relative to that of sample B)

36
First arrays did not look good...
37
Array 2
38
Array 3
39
Technique improvement...
40
(No Transcript)
41
Technique improvement...
Less background problems but also less signal
42
Spotfire analysis showed

• Technique needs improvements!
• Location of the antibodies on the Microarray
• Some high background antibodies
• Procedure
• Normalization method

43

Participants

• BiGCaT Bioinformatics
• Rachel van Haaften
• Arie van Erk
• Chris Evelo
• Florence University
• Christina Luceri
• Funding
• NuGO (exchange)
• NBIC (Spotfire server)